In short
- In the study, 70,000 applicants for customer service jobs were randomly interviewed by a human recruiter or an AI voice agent. Humans made every hiring decision.
- AI interview candidates were 12% more likely to get an offer, 18% more likely to start the job, and 17% more likely to still be employed four months later, with no drop in productivity.
- When they could choose, 78% of applicants picked the AI interviewer. The researchers credit structure: the AI covered the same topics, in the same order, every time.
- To prepare, read the job posting, answer every question fully with a real example, test your setup, and know your rights: New York City, for one, requires notice before an automated hiring tool is used.
On this page
What the AI interview study found
Economists Brian Jabarian (now at Carnegie Mellon, formerly at Chicago Booth) and Luca Henkel (Erasmus University Rotterdam) worked with a recruiting firm hiring for entry-level customer service jobs in the Philippines. Applicants were randomly assigned to a human recruiter, an AI voice agent, or a choice between the two. Candidates knew when they were talking to an AI, and in every case human recruiters reviewed the transcript, the audio and a standardized test score before deciding.
| Outcome | Interviewed by AI, compared with a human |
|---|---|
| Received a job offer | 12% more likely |
| Started the job | 18% more likely |
| Still employed after 2, 3 and 4 months | 17%, 16% and 17% more likely |
| Productivity once hired | No difference |
| Chose the AI when given the option | 78% of applicants |
| Reported gender-based discrimination | 3.30% with the AI, 5.98% with a human |
Recruiters did not see it coming: before the results, 48% of them expected workers hired this way to stay for less time. The AI was not flawless either. About 5% of applicants hung up because they did not want to speak to an AI, the voice agent ran into technical problems in about 7% of cases, and applicants rated the AI conversation as less natural.
What the study does not say
- It is one employer and one kind of job. The results come from high-volume, entry-level customer service hiring in the Philippines. They may not hold for a software engineer, a nurse or a manager hired in the United States.
- The AI did not decide. It asked the questions; people made the offers. A system that scores or rejects applicants on its own is a different product.
- It is a working paper. The paper has been updated during 2026 and is still a working paper. The authors disclose that Jabarian took an unpaid role with the partner firm, PSG Global Solutions, after the data were collected.
Why a structured AI interview can help candidates
"AI is not so magic," Jabarian told the Chicago Booth Review. Both the recruiters and the AI had the same list of up to 14 topics to cover. People drift: they skip topics late in the day, change the order, ask follow-ups unevenly. The AI interviewer covered more topics, in a more consistent order and wording, while still reacting to each answer. The researchers call it controlled variance, and say it gave recruiters more of the information they needed to judge each applicant.
For you, that means the conversation is less about charm and more about content. In the transcripts, a sustained back-and-forth with the interviewer went with more job offers. Every topic is likely to come up, so every answer counts.
What an AI interview looks like
The words cover a few different formats. Read the invitation carefully to know which one you face:
- A voice or video conversation with an AI interviewer, like the one in the study: it introduces itself as an AI, asks set questions and follow-ups, and records the call for a recruiter.
- A one-way video interview: questions appear on screen and you record your answers within a time limit. Software may transcribe or score them before a person watches.
- A chat screening: a chatbot on the career site asks about your availability, location, work authorization or experience before passing you on.
In all three, the job posting is the script. The questions come from the skills and duties it lists, and so should your examples.
Common AI interview questions to practice
An AI interviewer follows a script built by the hiring team, usually from the job description. Before your first AI interview, plan your prep around the questions most screening rounds include:
- Tell me about yourself: 60 to 90 seconds on your current role, one result, and why this job is the next step.
- Why this company and this role? Name something specific from the job description or the company's website.
- A behavioral question: "Tell me about a time you handled an upset customer" or "a deadline you almost missed". Use one real story.
- A skills question: the tools, software or certifications listed in the posting, and how you have used them at work.
- Logistics: availability, shifts, location, start date and, often, salary expectations. Have honest, short answers ready.
Practice out loud, not in your head: record yourself on your phone, listen back, and cut the parts that wander. Feedback from a friend, a career center or a practice tool helps you track progress and build confidence. Our guide on how to prepare for an interview covers the rest of the plan, from research to follow-up.
How to prepare for an AI interview
- Turn the job description into your list of interview questions. For each specific requirement, prepare one real example from your experience. The STAR method (situation, task, action, result) keeps each answer complete.
- Answer the question that was asked, fully. An AI interviewer will not read your face or fill the gaps for you. Give the context, what you did, and the result, in about one to two minutes.
- Prepare the classics. Expect "tell me about yourself", your reasons for applying and your availability. Our guide to tell me about yourself has examples.
- Test your setup. Use a quiet room, a charged phone or laptop and a stable connection. If the call drops or the AI stops responding, write down the time and email the recruiter right away.
- Speak clearly and keep your resume next to you. Dates, job titles and numbers should match what you sent.
- Do not use a hidden AI tool to answer for you. A human reads the transcript, and the answers have to hold up in the next round with a person.
Your rights in an AI interview in the United States
The rules depend on where you live. In New York City, Local Law 144 of 2021 bars employers and employment agencies from using an automated employment decision tool unless it has had a bias audit within the past year, the audit results are public, and candidates receive notice 10 business days before the tool is used. The city's Department of Consumer and Worker Protection takes complaints, and you do not need an account to file one.
Elsewhere, ask the recruiter how the tool is used: who reviews it, whether it scores you, and whether another format is possible if you need an accommodation. Software used to screen applicants is also being tested in court, as our page on the Workday lawsuit explains.
Frequently asked questions.
Sources
- AI Job Interviews Lead to Better Outcomes (November 13, 2026), Chicago Booth Review
- Voice AI in Firms: A Natural Field Experiment on Automated Job Interviews, Brian Jabarian and Luca Henkel (working paper, version of January 26, 2026), CIRJE, University of Tokyo
- Voice AI in Firms: A Natural Field Experiment on Automated Job Interviews (CESifo Working Paper No. 12984, September 10, 2026), CESifo
- Automated Employment Decision Tools (AEDT), NYC Department of Consumer and Worker Protection

About the author
Diego Benitez
Founder of Kandiq, more than 100 hires led
Diego Benitez has hired more than a hundred people, in marketing, sport and sales, as a manager at Intersport and then as Head of Growth at Goodvest (2022 to 2026). He ran those hiring processes from start to finish: screening applications, interviews, the hiring decision, then managing the team. He also knows the candidate's side: internships, work-study, student jobs, a permanent contract, starting an agency, then freelancing. He built Kandiq to give candidates what he looked for in an application himself, and what screening software can read. His guides start from the questions candidates ask Google, rely on official public sources (government employment and careers services, employment law) and are updated when the rules change. The examples are fictional; the figures are sourced.

